On Hardware Solution of Dense Linear Systems via Gauss-Jordan Elimination

Abstract

Gauss-Jordan Elimination (GJE) is a popular method for solving systems of linear equations. Much work has been done to design high throughput, low cost, FPGA-based architectures for GJE. However, as the interest in energy efficient designs increases, power consumption becomes a prevalent metric that must be considered in any FPGA-based implementation. In this paper, we present a scalable architecture that can efficiently solve any generic system of linear equations using GJE with a single-precision floating-point accuracy and reasonable power and area overheads. Comparisons with two previous implementations show the efficiency of our design.

Publication
In the 2015 IEEE Pacific Rim Conference on Communications, Computers and Signal Processing (PACRIM)

Related